The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates environment management by executing the zstack binary, including instructions for local builds using go build in a specific user directory.
[COMMAND_EXECUTION]: Provides interactive access to managed environments via the zstack shell command.
[CREDENTIALS_UNSAFE]: Orchestrates authentication workflows using zstack auth login which processes email and password credentials, and supports the creation of scoped API tokens.
[DATA_EXFILTRATION]: Exposes high-privilege data access commands including zstack secrets list, zstack files download for retrieving environment files, and zstack db query for executing arbitrary SQL.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by interpolating untrusted parameters such as environment IDs, branch names, and hostnames into shell commands. Ingestion points: User-provided arguments for environment, branch, and domain commands. Boundary markers: None present. Capability inventory: Runtime shell access, database queries, file downloads, and secret listing. Sanitization: None described; safety depends on the underlying CLI implementation.

zstack-cli